Curriculum Modulesmodule-15Vector Database Architecture (Qdrant, Milvus, pgvector)

Vector Database Architecture

Introduction

Vector databases are specialized data stores optimized for storing and querying high-dimensional vectors. In the context of RAG, they serve as the memory engine for LLMs.

Popular Vector Databases

  • Qdrant: Rust-based, highly performant, API-first vector search engine.
  • Milvus: Cloud-native vector database built for massive scale.
  • pgvector: PostgreSQL extension adding vector similarity search.

Deployment Strategies

We explore deploying these on Kubernetes using Helm, setting up replication, and managing persistent volumes for vector storage.

Designing Retrieval-Augmented Generation (RAG) InfrastructureAutomated Model Weight Ingestion & Storage Pipelines